Functional genomics in Inborn errors of immunity

Summary

Inborn errors of immunity (IEI), comprise a diverse spectrum of 485 disorders as recognized by the International Union of Immunological Societies Committee on Inborn Error of Immunity in 2022. While IEI are monogenic by definition, they illuminate various pathways involved in the pathogenesis of polygenic immune dysregulation as in autoimmune or autoinflammatory syndromes, or in more common infectious diseases that may not have a significant genetic basis. Rapid improvement in genomic technologies has been the main driver of the accelerated rate of discovery of IEI and has led to the development of innovative treatment strategies. In this review, we will explore various facets of IEI, delving into the distinctions between PIDD and PIRD. We will examine how Mendelian inheritance patterns contribute to these disorders and discuss advancements in functional genomics that aid in characterizing new IEI. Additionally, we will explore how emerging genomic tools help to characterize new IEI as well as how they are paving the way for innovative treatment approaches for managing and potentially curing these complex immune conditions.

Introduction

Inborn errors of immunity (IEI)¹, comprise a diverse spectrum of 485 disorders as recognized by the International Union of Immunological Societies Committee on Inborn Error of Immunity in 2022². As the prior nomenclature indicated, they were historically considered primary immunodeficiencies due to their association with heightened infection susceptibility. As exponential numbers of diseases have been described, it has been recognized that IEI are primarily Mendelian disorders driven by distinct, often rare, coding mutations that disrupt immune function. These mutations result in increased susceptibility to recurring and atypical infections, but also autoimmunity, cancer, and other abnormal immune responses. IEI are now often subdivided into two main categories: primary immunodeficiency disorders (PIDD), in which the central problem is heightened susceptibility to infection, and primary immune regulatory disorders (PIRD), in which the central problem is immune-mediated (i.e. autoinflammatory/autoimmune) manifestations³. Intriguingly, mutations in the same gene can result in dramatically different phenotypes, specifically leading to either PIRD or PIDD phenotypes: in some extreme cases, gain of function mutation of one gene can lead to a PIRD phenotype, while loss of function mutation of the same gene leads to a PIDD⁴. This phenomenon illustrates the critical role of certain genes in maintaining immune homeostasis and underscores how different types of mutations can tip the balance in opposing directions. As an example of this complex relationship, TLR7 gain of function mutations are associated with systemic lupus erythematosus (SLE)^5,6 while hypomorphic variants of TLR7 can result in severe viral infections⁷. In some other cases, mutations in a single gene can lead to complex clinical presentations that blend features of both autoimmunity and immunodeficiency. This is the case of Autoimmune Polyglandular Syndrome Type 1 (APS1), also known as Autoimmune Polyendocrinopathy-Candidiasis-Ectodermal Dystrophy (APECED). APS1 is caused by mutations in the AIRE (Autoimmune Regulator) gene, which plays a critical role in establishing central immune tolerance, a mechanism that leads to elimination of self-reactive T cells⁸. Patients with APS1 predominantly present with a range of autoimmune symptoms affecting multiple endocrine glands, but they also frequently suffer from chronic mucocutaneous candidiasis, indicating an increased susceptibility to certain types of infections.

While IEI are monogenic by definition, they illuminate various pathways involved in the pathogenesis of polygenic immune dysregulation as in autoimmune or autoinflammatory syndromes, or in more common infectious diseases that may not have a significant genetic basis⁹. Advancements in our understanding of IEI have profound implications beyond these rare conditions, and can help unravel the fundamental mechanisms of the human immune system. Indeed, several genes implicated in IEI also play a role in more common autoimmune diseases, offering a glimpse into the shared genetic architecture of these disorders⁹. Mutations in NOD2 can lead to Blau syndrome, an IEI characterized by granulomatous arthritis, uveitis, and rash¹⁰. Variants of this gene are also involved in more common conditions like Crohn’s disease¹¹. Understanding these nuanced genetic determinants can yield significant insights into immune regulation and open avenues for targeted therapies tailored to the specific nature of the mutation and the resultant immune disorder. Through allelic series in genes like NOD2, we can trace a connection between rare genetic errors and widespread autoimmune or infectious disorders, shedding light on shared pathways of immune dysregulation. They offer valuable insights into the complex interplay of genetic factors that contribute to immune dysfunction across a spectrum of diseases, paving the way for cross-disciplinary research and potential therapeutic interventions.

Rapid improvement in genomic technologies has been the main driver of the accelerated rate of discovery of IEI and has led to the development of innovative treatment strategies¹². In this review, we will explore various facets of IEI, delving into the distinctions between PIDD and PIRD. We will examine how Mendelian inheritance patterns contribute to these disorders and discuss advancements in functional genomics that aid in characterizing new IEI. Additionally, we will explore how emerging genomic tools help to characterize new IEI as well as how they are paving the way for innovative treatment approaches for managing and potentially curing these complex immune conditions.

Primary immunodeficiency disorders (PIDD) and infectious disease.

The first clearly recognized IEI were reported in the 1950s and included several prototypical but mechanistically distinct PIDD including agammaglobulinemia¹³, severe combined immunodeficiency¹⁴, and chronic granulomatous disease¹⁵. Over the subsequent decades, these disorders were better described, additional ones identified, and what were initially viewed as a single disease entities subsequently recognized to have locus heterogeneity, at times with associated distinct clinical features¹⁶. As the rare, highly penetrant mutations underlying IEI have been identified and the associated pattern of increased infection susceptibility defined, the search for common variants affecting disease susceptibility has had more modest results. GWAS studies have identified associations for various infectious diseases, including hepatitis B virus, tuberculosis, and Helicobacter pylori infection, though the overall yield has been less striking than in the study of a number of chronic non-infectious diseases¹⁷. Most recently, in COVID-19, GWAS of 49,562 cases and over 2 million controls identified genetic variants associated with disease, many of which are genes implicated in type-1 interferon signaling¹⁸. The overall heritability of COVID-19, similar to other infectious diseases except for leprosy, was between 1–6.5%^18–19. Other observations during the COVID-19 pandemic, however, brought into sharp focus the relationship between PIDD and prevalent infectious diseases. Early observations made in 2020 revealed that approximately 10–20% of patients enduring severe to critical COVID-19, necessitating mechanical ventilation, exhibited high concentration of autoantibodies (autoAbs) neutralizing type 1 interferons (IFN) in their serum²⁰. These specific type 1 IFN autoAbs were first identified in patients with mutations in the RAG1 and RAG2 genes²¹. It was also realized during the pandemic that patients with APS1 developed severe cases of COVID-19, which was directly related with their increased risk of having pre-existing anti-type 1 IFN autoAbs²⁰. Here it is observed that the phenotype contributing to heightened susceptibility to COVID-19 in APS1 patients was phenocopied, in a non-Mendelian manner, in a much larger portion of the population, who exhibited comparable increased risk. This observation resulted in the inclusion of individuals with autoAbs against type 1 IFNs as the first non-genetic condition in the International Union of Immunological Societies Committee on Inborn Error of Immunity in 2022².

Primary immune regulatory disorders (PIRD) and common autoimmunity.

While PIDD were the first IEI described, and IEI associated with infections make up the majority of IEI, there is a growing number of IEI with clinical manifestations resulting from immune-mediated pathology, leading to autoimmunity, autoinflammation, atopy, or lymphoproliferation and malignancy. This group of disorders, PIRD, comprises multiple phenotypes sharing features with common polygenic rheumatologic diseases. This overlap suggests that common autoimmune diseases may share underlying immunological pathologies with PIRD and that mutations causing PIRD can inform the causes of more common autoimmune disease. Mutations in the ADA2 gene can result in Deficiency of Adenosine Deaminase 2 (DADA2), a disease characterized by recurrent fevers, livedo reticularis, and early-onset strokes²². Interestingly, there has been an increasing recognition that some cases of Polyarteritis Nodosa (PAN) are in fact also due to mutations in ADA2²³. This discovery highlights the complex and sometimes overlapping clinical presentations that can arise from mutations in a single gene, further emphasizing the need for genetic evaluation in understanding the full spectrum of immune disorders.

The clinical manifestations of PIRD often have significant overlap with common autoimmune diseases with reduced and polygenic heritability. SLE is a complex systemic autoimmune disease that is considered to result from a combination of genetic predisposition as well as environmental factors such as UV radiation, infections and hormones. The genetic association is supported by a concordance rate of lupus in monozygotic twins of up to 25% as compared to 2–7% in dizygotic twins²⁴, and multiple loci have been reported to be associated with increased susceptibility²⁵. In addition to those cases, there are also multiple well-established cases of monogenic lupus, a form of SLE that usually presents early in life and is typically more severe. Typically, the cases of monogenic lupus do not exhibit all the classic characteristics of adult-onset disease, but there are enough similarities between those diseases to make them a useful tool to understand important pathways in the pathogenesis of lupus, and guide potential treatments²⁶. For example, interferonopathies such as Aicardi-Goutieres syndrome share common features with SLE, such as chilblains, rashes, arthritis, oral ulcers and autoantibodies such as antinuclear antibodies (ANA) or anti-dsDNA antibodies. Similarly, multiple monogenic forms of SLE are related to complement deficiencies, which lead to defects in the clearance of apoptotic cells and immune complexes²⁷. For example, mutations in the C1Q genes (C1QA, C1QB, C1QC) leading to deficiency in C1Q are associated with rashes, mucosal ulcers, arthritis and nephritis, and interestingly anti-C1q antibodies is a common autoantibody found in SLE patients with nephritis²⁸. Overall, this serves as a proof of concept of how understanding the role of genes identified in monogenic IEI can enhance our knowledge of pathways and molecular mechanisms involved in more common autoimmune diseases. As another example, specific loss-of-function mutations in TNFAIP3 can lead to A20 haploinsufficiency, inducing an autoinflammatory disease that resembles Behcet’s and SLE²⁹. Additionally, both common missense variants of TNFAIP3 and non-coding variants influencing its expression have been associated with SLE in large population genome-wide association studies (GWAS)³⁰. The manifestation of similar symptoms across autoimmune disease strongly highlights that the significance of IEI extends far beyond the scope of rare disease³¹.

Mendelian inheritance, human population structure, genetic architecture

Each individual harbors millions of mutations³², many of which are exceedingly rare in the general population^33–34. For example, in the Trans-Omics for Precision Medicine (TOPMed) program across 53,831 TOPMed samples, there were 400 million single-nucleotide and insertions of which 97% have a frequency less than 1% and 46% are present in one individual. The rarity of these mutations is partially due to the population bottlenecks resulting from multiple emigration events out of Africa; and the lack of genetic diversity in even the latest genome sequencing efforts. Variants that cause IEI, like other Mendelian diseases, are highly enriched among the rarest variants. While the first IEI were identified by linkage studies, next generation sequencing has transformed our ability to identify rare variants. Nevertheless, some coding variants for IEI remain unknown and given their rarity, may be difficult to identify through traditional linkage and association analyses of human populations³⁵. Identifying causal non-coding variants associated with IEI may be even more challenging. Indeed, like other Mendelian disorders, many individuals who present symptoms of IEI may be among the ~50% who do not have a definitive genetic diagnosis.

Mendelian inheritance follows the principles of segregation and independent assortment. These principles hold that individual alleles are passed from parents to offspring in specific ratios, determining various molecular and physiological phenotypes. Most IEI follow Mendelian inheritance and result from germline variants in single genes and are often coding in nature. The rationale behind this is based on the functional role of coding regions in the genome. Coding regions, or exons, are parts of the genome transcribed into mRNA and then translated into proteins, which are vital to the body’s structure, function, and regulation of biological processes. Therefore, alterations in these coding regions, even minor ones, can have a profound impact on the protein’s function, potentially leading to disease. Non-coding regions, on the other hand, mainly regulate gene expression, and alterations here may not have as immediate or severe an effect as in coding regions. Thus, coding mutations are more likely to manifest as phenotypically identifiable, and often rare, Mendelian disorders. The disease-causing variants often result in loss-of-expression, loss-of-function (LOF; amorphic/hypomorphic), or gain-of-function (GOF; hypermorphic) of the encoded protein. Inheritance of a single variant may present as an autosomal dominant pattern of inheritance caused by GOF, haploinsufficiency, or negative dominance. Biallelic inheritance typically presents an autosomal recessive pattern of inheritance by LOF of the encoded protein (rarely GOF), while X-linked recessive traits arise from LOF of genes on the X chromosome.

In contrast, common autoimmune diseases like Crohn’s disease, Behçet’s disease, and SLE are polygenic, resulting from a complex interplay of multiple genetic variants and environmental influences. While each genetic variant might contribute a modest risk, their cumulative effect significantly influences disease predisposition^36–37. Despite their different genetic origins, recent research has revealed intriguing links between Mendelian disorders like IEI and common autoimmune diseases. For instance, specific loss-of-function mutations in the TNFAIP3 gene, implicated in certain IEI, can lead to A20 haploinsufficiency, a condition that triggers autoinflammatory diseases with features resembling Behçet’s disease and SLE²⁹. Similarly, PFAPA (periodic fever with aphtous stomatitis, pharyngitis and adenitis) is the most common periodic fever syndrome in children. Interestingly, PFAPA, Behçet’s disease and more common recurrent aphthous stomatitis (aka canker sores) share common genetic variants, placing them on a continuum of diseases from the more rare end of IEI to much more frequent conditions^38–39. As a third example, NOD2 coding mutations, originally identified in the context of IEI, have been linked to an increased susceptibility to Crohn’s disease⁴⁰. Additionally, GWASs have highlighted common variants in these genes as risk factors for these autoimmune diseases⁴¹. These examples illustrate that the genetic architecture of Mendelian disorders like IEI and common autoimmune diseases can reside on a shared continuum, with the same genetic pathways often implicated in both. As such, insights from studying Mendelian disorders like IEI could provide valuable knowledge into the genetic basis and pathophysiology of common autoimmune diseases, and vice versa. Understanding this connection represents a critical aspect of advancing genetic and genomic research.

Interestingly, unlike traditional germline mutations that manifest early in life in many IEI, recent diseases have been shown to be due to somatic mutations and typically affect adults. This is the case of Vacuoles, E1 enzyme, X-linked, autoinflammatory, somatic syndrome (VEXAS), a recently described inflammatory disease primarily affecting adult males, that has broad manifestations including fevers, cytopenias, features of vasculitis, chondritis or neutrophilic syndromes among others⁴². It is due to a somatic mutation in the UBA1 gene, which encodes a ubiquitin-activating enzyme essential for protein degradation. The integration of VEXAS into the IEI landscape underscores the complexity and expanding scope of inborn errors of immunity, which now includes diseases caused by somatic mutations in addition to those caused by germline mutations. The identification of VEXAS syndrome stands as a prime example of how cutting-edge genomics techniques can aid in the discovery of previously unknown conditions⁴³.

What is functional genomics today

Functional genomics aims to understand the relationship between an organism’s genome and its phenotype. The term first arose in the scientific literature in the late 1990s as high-resolution genome sequences became available. A 1997 review of the field defined it as, “the development and application of global (genome-wide or system-wide) experimental approaches to assess gene function by making use of the information and reagents provided by structural genomics⁴⁴.” Phenotypes are usually hierarchical, with disease manifestations emerging from tissue and organ phenotypes, which arise from cellular phenotypes in which specific biological pathways are altered, which in turn reflect molecular phenotypes such as altered chromatin accessibility, gene expression, or protein function. Determining correlative and causative relationships between genotypes and phenotypes requires variation in both.

Many advances in functional genomics are built on using sequencing as an approach to profile phenotypes across multiple levels of biological function. This is both through converting non-DNA signals into DNA-based ones and through the use of DNA molecules as a barcoding system, enabling massively parallel and highly multiplexed experiments. The most common functional genomic assay aims to profile transcript abundance by reverse transcribing RNA transcripts to create cDNA fragments flanked by sequencing primers that are then quantitated by next-generation sequencing^45–46. To assay DNA methylation, bisulfite treatment is used to convert unmethylated cytosines into uracils while methylated cytosines remain unchanged; through sequencing, the degree of methylation can be estimated by identification cytosine to thymine conversion^47–48. Histone modifications such as acetylation and methylation and binding of specific transcription factors are assayed with ChIP-seq (chromatin immunoprecipitation and sequencing)⁴⁹. In ChIP-seq, DNA is chemically cross-linked to nearby proteins, the transcription factor or modified histone of interest is bound and pulled down using a specific antibody, and the DNA then dissociated and sequenced⁵⁰. In ATAC-seq (assay for transposase-accessible chromatin using sequencing), a hyperactive Tn5 transposase enables profiling of regions of accessible chromatin with very little input material⁵¹. Chromosome conformation capture provides insight into 3-D chromatin structure by crosslinking regions of interacting DNA followed by next generation sequencing⁵². These approaches have now been applied to a large number of human cell types and tissues and collected together for public reference as ENCODE (the Encyclopedia of DNA Elements)^53–54. Beyond labeling individual molecular outputs, recent advances have also used DNA molecules as barcodes for biological contexts. They allow for cell-specific labeling in single-cell RNA-sequencing, which can be performed in a high-throughput fashion using droplet-based approaches^55–56. Use of barcoded antibodies to proteins of interest allows for protein expression quantitation concurrently with RNA-seq (e.g. CITE-seq⁵⁷ or REAP-seq⁵⁸). In addition, there are a number of approaches that use DNA molecules to barcode spatial locations⁵⁹.

Collectively, the above approaches allow the simultaneous profiling of numerous molecules across cellular contexts and modalities⁶⁰. The resulting data can in turn be analyzed to determine how different functional molecules interact as a system to drive biological processes and respond to environmental changes or disease development. Importantly, these advances also enable the use of individual compartments as units of experimentation. For example, single-cell RNA-seq or multimodal analysis allows for useful analysis of cells that are individually and variably genetically perturbed whether by natural or induced genetic variation. The effects of the genetic perturbation on multiple cellular phenotypes can be simultaneously measured, providing an unprecedented opportunity to examine the effects of variants across multiple scales. By understanding these gene functions and interactions, functional genomics has the potential to inform us about the genetic basis of diseases and potential therapeutic interventions. It is also playing a crucial role in the development of precision medicine, in which treatments are tailored to an individual’s specific genetic profile.

Forward, genetics: Coding variant functionalization

Coding variants–that is, variants occurring in the protein-encoding portions of the genome–provide important insights into protein function as well as population genetics and evolutionary biology. Nonsynonymous variants, including substitutions, frameshifts, and premature stop codons, have a straightforward effect on the amino acid sequence; providing an easier entry point into their study than non-coding variants and providing a pragmatic explanation for the greater efforts on their characterization thus far. Understanding the broader effects of a given amino acid change is of course not nearly as straightforward and has supplied ample opportunity for decades of investigation. Although most research has focused on non-synonymous variants, evidence has been found that synonymous variants may also result in altered phenotypes, at least in certain settings^61–62. Significant recent advances have been made in both in silico analyses that aim to predict the effect of coding variants and experimental approaches that permit study of variants in a high-throughput fashion.

From a population genetics perspective, harmful coding variants are often subject to negative selection, meaning they are likely to be eliminated from a population over time because they decrease the fitness of individuals carrying them. Consequently, coding variants that persist in a population and cause diseases like IEI are often of great interest, as their existence suggests they may have compensatory advantages or could be simply recent mutations or the result of genetic drift. From an evolutionary standpoint, the study of coding variants in IEI can shed light on the balancing act between an effective immune response and the risk of autoimmunity. The immune system must walk a fine line, swiftly recognizing and targeting foreign pathogens while not reacting to the body’s own cells. The genetic variants that are implicated in IEI offer valuable insights into the essential components of the immune system, as well as the genetic diversity and evolutionary pressures that shape it. There are enormous opportunities to functionalize coding variants (i.e., determine the effect of a coding variant on protein function), which could have profound effects on our ability to diagnose and treat patients with IEI.

There is a more than two-decade-long lineage of computational tools that aim to predict the functional consequences of a given coding variant. Recent innovations in deep learning have pushed the performance of such methods forward. Specifically, the use of protein language models, which are derived from natural language processing methods, have been at the center of these advances^63–67. These neural networks are trained on protein sequences selected from across the domains of life (e.g. ~250 million sequences in UniProt⁶⁴). Although the input training data are amino acid sequences alone, without any clinical or functional annotations, these models implicitly learn information about protein structure and function. In a recent study, we applied evolutionary-scale models (ESM), a protein language model developed by MetaAI, to predict the effects of all possible missense mutations in the human genome⁶⁸ (Fig. 1A). We observed an area under the curve of 0.91 in predicting benign versus pathogenic variants in ClinVar or between variants segregating in gnomAD or annotated in HGMD. ESM’s performance is particularly impressive at regions of the receiver operating characteristic curve corresponding to a low false positive rate, which may be most relevant in a clinical diagnosis setting. ESM scores are also highly correlated with molecular phenotypes as measured in saturation mutagenesis screens. ESM, together with related methods such as EVE⁶⁹, primate AI⁷⁰, and AlphaMissense⁶⁷, have shown themselves to be the current first-in-class in silico methods for predicting the function of coding variants^68–70. However, despite their high accuracy, protein language models provide limited interpretation of the consequences of mutations, wherein the analysis yields a single score that essentially predicts the divergence of a mutation from the wildtype (WT) sequence. Deep language models implicitly learn protein structure utilizing a high dimensional latent space, which is then converted into that single score. However, the actual high-dimensional latent space might hold even more information about variant features.

Figure 1. Advancing Coding Variant Functionalization Through Computational and Experimental Genomics.

(A) Utilizing Protein Language Models for Coding Variant Prediction. Pretrained ESM1b neural network is applied to an amino acid sequence to predict the functional impact of all possible amino acid changes as illustrated with the heatmap of LLR effect scores. This method has a high predictive value of coding variant effects, distinguishing benign from pathogenic mutations with an area under the curve (AUC) of 0.91 when compared to ClinVar.

(B) High-Throughput Coding Variant Functional Assays with Perturb-seq. Cells from a cell line are transduced with a library of variants, each linked to a unique barcode and are sequenced. This enables the parallel assessment of the impact of each variant on cellular behavior and transcriptional profiles by comparison with cells overexpressing the WT version.

(C) CRISPR-Select for Targeted Variant Analysis. Cells are transfected with a CRISPR-Select cassette including variants of interest and synonymous internal normalization mutation single-stranded oligodeoxynucleotide (ssODN) repair templates. Comparison of variant:WT′ ratios (1) at an early and subsequent timepoint allows determination of variant effect on proliferation, survival or fitness (CRISPR-Select TIME); (2) in an initial compartment and a spatially distant compartment allows determination of variant effect on cell motility, invasiveness and homing; (3) in high versus low expression of cell state markers as measured by flow cytometry allows determination of variant effect on differentiation, metabolism and cell function.

Experimentally, the power to produce genetic variants, either with synthetic DNA sequences or direct endogenous genome editing, has opened up exciting avenues to functionalize coding variants in biochemical or cell-based assays. These Perturb-seq-like studies simultaneously profile the effects of multiple variants in a given gene; initial versions of these assays focused on deciphering a single read out^71–72, with later studies expanding to sequencing the entire transcriptome to discern more complex transcriptional phenotypes. An example of the latter is the recent study by Ursu et al., which examined missense mutations in TP53 and KRAS by transfecting cells with an exogenous copy of the gene⁷³ (Fig. 1B). The promoter of the exogenous copy is designed to dramatically dominate the endogenous expression of the gene (e.g. >10 fold). By comparing the transcriptional profiles between cells transfected with a mutated copy of the gene and cells transfected with the WT copy, the functional consequences of mutations can be determined. If cells with a mutation are indistinguishable from cells with a WT copy, the mutation is determined to be WT-like/neutral. If they are indistinguishable from untransfected cells carrying only the endogenous copies of the gene, the mutation is determined to be loss-of-function. Otherwise, the mutation is determined to be “change of function” (e.g. gain of function, dominant negative)⁷³.

Introduction of a library of variants using lentiviral transduction, as used in Ursu et al, has been an effective approach but comes with certain limitations, such as issues with multiplicity of infection and potentially high variability in the expression of the transgene depending on the site of gene integration within the host cell genome. Alternative strategies have focused on ensuring the exogenous copy is introduced into a consistent location in the genome. One such approach has been to engineer a “landing pad” site into cells, in which there is a Tet-inducible cassette with a Bxb1 recombination site, which can accommodate a single copy of gene, permitting each successfully transduced cell to have a single variant of interest⁷⁴. Improved iterations of this idea have more recently described⁷⁵. Applications of this approach include deep mutational scan of the SCN5A voltage sensor, characterization of the effect of thousands of variants on expression levels of PTEN and TPMT⁷⁶, and study of variant impacts on stability and function of VKOR⁷⁷. Such an approach could be harnessed to explore genes important in immunologic function, such as deciphering the effect of variants on surface receptors crucial for the communication between immune cells and their environment.

Study of variants using an exogenous version of the gene of interest may avoid interference by the endogenous gene through any of a few approaches, including use of a highly active promoter to produce much higher expression of the introduced copy of the gene, introduction of the gene into a cell that does not usually express the gene of interest, or by using cells in which there is prior disruption of the endogenous copy of the gene (e.g. by CRISPR-targeted knockout). A complementary line of research, however, sidesteps this problem by introducing variants directly into the endogenous gene. In particular, in recent years there has been a progression from the targeted introduction of a small number of variants to highly parallel profiling of numerous variants. One approach has been to use CRISPR-Cas9 with homology-directed repair (HDR), for example to perform near-saturation mutagenesis of critical exons of BRCA1⁷⁸. Niu et al have recently described CRISPR-Select, a system for variant profiling across a variety of readouts, including proliferation/survival, migration, and marker expression, with precise quantitation through use of an internal, control mutation⁷⁹ (Fig. 1C); although as described the throughput of this approach may be lower than some alternative techniques, its advantages in the breadth and specificity of its results indicate utility for certain questions.

Another approach for these variant screens has been to use base editors. Dead or nickase modifications of the Cas9 enzyme are coupled to either a cytosine base editor, which convert a cytosine to a uracil, ultimately resulting in a thymine in that position, or an adenine base editors, which convert adenine to hypoxanthine, ultimately resulting in a guanine in that position⁸⁰. Since their original development just several years ago, multiple generations of improvements to these base editors have been developed, resulting in improved efficacy, specificity, and range of targetable sites^81–85. Highly parallel variant profiling using base editing has been used in a number of cases already. These include evaluating more than 50,000 ClinVar variants across more than 3000 genes⁸⁶, identifying loss-of-function and gain-of-function variants in multiple elements of the DNA damage response⁸⁷, and a mutagenesis screen across hematopoietic transcription factor GATA1⁸⁸.

The choice of means by which the function of a variant is assessed has important implications for the genes in which the approach is likely to be useful. For example, changes in frequency over the course of cycles of proliferation are useful for genes affecting proliferation or survival; comparison of transcriptomes will be most useful for that directly affect transcription (i.e., transcription factors) or those that have substantial downstream transcriptional effect (e.g., signaling molecules). Broadening the set of examined effects can therefore increase potentially what can be learned from these variant screens. For example, combining perturbations with chromatin accessibility assays could allow us to potentially functionalize transcription factors by directly assessing their binding or other chromatin remodelers. Other approaches, including those that enable scalable proteomic measurements such as SCITO-seq, a modification of conventional CITE-seq permitting high-throughput⁸⁹, allows functionalization of coding variants to examine their effect on protein folding and trafficking to the surface. Indeed, additional approaches to profile using sequencing phosphorylation events or changes in ion gradients could enable multiplexed experiments to differentiate the effects of coding variants on protein folding versus their catalytic effects.

Non-coding variant functionalization

Identifying non-coding genetic variants that cause IEI and other Mendelian diseases and unveiling their functionality marks a crucial progression in our understanding of genomic complexities that cause human disease. Non-coding variants pose a unique set of challenges and opportunities. They necessitate a shift from the starting point of predictable amino acid consequences as can be determined from coding variants. Another challenge is at the data analysis step. The identification of non-coding variants requires whole genome sequencing (WGS). While the experimental cost for WGS has significantly decreased, WGS generates ~90% more data than whole exome sequencing (WES) and thus analyses of WGS data remains a major technical hurdle. Alternative approaches that only sequence selective regions of the genome near genes may be an attractive alternative to focus resources to regions where causal variants are most likely to reside.

As with coding variants, non-coding variants with strongly deleterious effects will likely be rare secondary to negative selection. However, the identification of rare variants with high penetrance or strong effect may be informed by the presence of common variants. Genome-wide association studies (GWAS) of multiple relatively common autoimmune diseases have linked hundreds of genes to these disorders^90,91. These studies have presented both the challenge and opportunity of annotating the specific causal common variants and the nature of the effects and suggested sites in which focused investigation for more highly penetrant alleles may be productive.

Genetic analysis of common forms of autoimmune diseases have identified hundreds of genes harboring common variants associated with these diseases. These variants are overwhelmingly non-coding and are hypothesized to mostly affect transcription. A major challenge in human genetics has been to annotate the effects of these variants including the cellular context by which they may act, what has now been appreciated as an important factor in determining how a variant ultimately affects disease susceptibility. For rare variants that cause IEI, we will inevitably be sequencing very few samples. In these cases, examination of allele specific effects will be necessary. In the case of RNA-sequencing, this would require that the causal variant is in direct linkage with an exonic variant.

One approach to functionalize non-coding variants is to perform biomolecular assays on large population cohorts utilizing natural genetic variation. Historically, this has been mostly pursued by measuring gene expression utilizing microarrays, Nanostring, or RNA-sequencing of bulk cells (for example)^92–94. What we and others demonstrated was that many common genetic variants control gene expression in specific cell types and in response to extracellular perturbations^95,96,97. In other words, the cellular context is essential for accurately annotating the effects of genetic variants. More recently, we have pioneered methods to utilize multiplexed single-cell sequencing as an approach to assess the effects of natural variants on gene expression across multiple cellular contexts simultaneously. Utilizing this approach, several large cohorts have now emerged to examine the effects of natural genetic variants on gene expression. We have recently used mux-seq to profile over 1.2M cells from 250 patients with lupus and 160 healthy controls⁹⁸. We identified that the depletion of naive CD4 T cells is inversely correlated with the induction of type-1 interferon stimulated genes in myeloid cells. By integrating dense genotyping data, we also mapped genetic variants associated with gene expression across 10 major cell types in circulation (Fig. 2A). This analysis confirmed our previous observations that genetic variants associated with SLE are much more likely to affect gene expression in specific cell types⁹⁹.

Figure 2. Advancing Non-coding Variant Functionalization Through Experimental Genomics.

(A) Characterization of Naturally Occurring Non-coding Variants. Cells from a human cohort undergo single-cell RNA sequencing (scRNA-seq), with cell-type assignment based on RNA-expression followed by identification of SNPs affecting gene expression and determination of whether this is in a global or cell-type-specific fashion.

(B) Non-transcriptomic Characterization of Non-coding Variants. Cells from a human cohort undergo ATAC-seq in addition to RNA-seq. Associations can be made between between common SNPs and gene expression (expression QTLs) and between common SNPs and regions of open chromatin (ATAC-QTLs).

(C) Massively Parallel Report Assays of cis-Regulatory Elements (CREs). A library of CREs with variants, a minimal promoter, a reporter gene, and a barcode specific to each variant, is transduced into a cell population. The effect of each variant on CRE activity is subsequently determined by measurement of reporter transcript (linked to the variant by specific barcode) normalized to the number of cells bearing that variant (measured by DNA sequencing).

(D) Variant Introduction and Characterization with Base Editing. A library of sgRNAs targeting sites of interest is introduced, together with a Cas9-based base editor, into a cell population of interest. Multiple analyses are possible, including transcriptomic studies using scRNA-seq and functional readouts in which variant enrichment or depletion in a subpopulation of interest is observed.

Beyond RNA, profiling more directly the biochemical activity of non-coding sequences might be a better approach to functionalize rare variants in small cohorts. For example, by measuring chromatin accessibility, to directly establish if a rare variant may affect accessibility. Other readouts, such as methylation, acetylation and DNase hypersensitivity can potentially be also adopted here though ATAC-seq has an advantage in its ease of use and low cell number requirements. Gate et al. demonstrated the promise of such an approach; they performed ATAC-seq and RNA-seq on primary T cells from over 100 donors and were able to identify natural non-coding variants that affect genome accessibility and gene expression¹⁰⁰.

Similar strategies could be developed to examine the effects of perturbations, not only natural variants. We and others have also started to perform multiplexed single-cell multiomics using the same approach to allow us to potentially assess rare disease-causing variants that affect chromatin state directly (Fig. 2B). Historically, systematic functionalizing of induced non-coding variants were performed by transducing cis-regulatory element sequence, often hundreds of base pairs long, with a minimal promoters and a reporter gene¹⁰¹. The reporter gene can be read out utilizing imaging and cell sorting but that is generally low throughput. Recent advances to encode each non-coding variant utilizing a barcode that is encoded in the transcript of the reporter gene has enabled highly multiplexed massively parallel reporter assays (MPRAs)^102–104 (Fig. 2C). However, MPRAs can be challenging to perform in primary cells and may be susceptible to experimental variability including the transduction approach. But most importantly, the effects of non-coding variants are likely affected by the genomic context and thus MPRAs do not reflect the native context where the sequences perform their function.

An alternative approach would be to use CRISPR genome engineering to perturb the genome. CRISPR based cutting, activation, inhibition and base editing each have their advantages and disadvantages¹⁰⁵. We have recently utilized CRISPR activation and inhibition to examine the non-coding variants at the CTLA4 locus where non-coding variants have been associated with rheumatoid arthritis. Recent efforts have demonstrated the utility of recapitulating large numbers of known non-coding variants in parallel. For example, Sharon et al developed CRISPEY, an approach that couples CRISR-Cas9 with bacterial retrons to allow massively parallel endogenous editing and variant functionalization. Applying this approach to S. cereviseae to look at more than 16,000 naturally occurring variants, they were able to identify hundreds of non-coding variants that affect fitness in glucose media¹⁰⁶. Base editors are particularly promising since they can endogenously edit the genome (Fig. 2D). The base editing approaches used to explore coding variants, as described above, are also being applied to non-coding variants⁸⁸. These editors are now able to make almost all possible edits and have the advantage of being compatible with highly multiplexed and massively parallel readout assays¹⁰⁷. However, there are still limitations for base editors, including that multiple different variants may be introduced by a single guide and not all mutations are targetable or can be targeted at the same editing efficiency^10,86.

Future opportunities for basic understanding

Genetic architecture

Even once a causal variant is identified, the consequences of the variant in a given individual are not always predictable. When the trait likely has a Mendelian pattern of inheritance, this discrepancy is encapsulated within the terms incomplete penetrance, where not all individuals carrying the disease variant express the disease, and variable expressivity, where the disease symptoms vary among individuals. For common traits that have a polygenic mode of inheritance, polygenic risk prediction (PRS) that models the collective effects of variants associated with disease have only modest ability to predict disease outcomes¹⁰⁸. The cause for this variability likely represents interplay between other genetic factors, environmental influences, and random biological events. Therefore, while identifying the causal variant is an essential first step, understanding its mode of action and its interaction with the host genome and the environment are all critical pieces of the puzzle of understanding the resulting phenotype. Studies in model organisms have allowed for controlled investigations of the role of genetic background on variant phenotype severity^109,110. Other research has built on this by identifying how coding variant penetrance is modified by cis-regulatory element variants¹¹¹. Whole genome sequencing of a cohort of individual with IEI similarly found interplay between high-penetrance rare variants and more common variants in the PTPN2 or SOCS1 loci¹¹². Such a comprehensive understanding is key in designing effective therapies, as it allows us to anticipate the variant’s behavior and the likely response to treatment in a more nuanced and individualized manner, consistent with the ultimate goal of personalized medical treatment.

Among genetically identically unicellular organisms there is substantial gene expression variability due to stochasticity, or randomness¹¹³. Such molecular variability within a clonal population can affect cellular response to signaling ligands or cancer cell resistance to therapy^114–115. Analyses in mice have similarly shown that even on a defined genetic background that incomplete penetrance and variable expressivity are common, illustrating the potential role of non-genetic factors including subtle environmental difference and stochastic events¹¹⁶. Similarly, a study of monozygotic twins discordant for multiple sclerosis could not identify genetic or epigenetic differences to explain the discordances¹¹⁷. Studies in human cell lines and model organisms have shown that deleterious variants in genes with paralogs, particularly if the paralogs have similar patterns of expression, can more often be compensated in comparison to genes without such paralogs^118–120. This phenomenon is anticipated to be active in humans as well.

Genetic interactions and stochastic factors may play a particularly significant role in IEI. This is perhaps best illustrated in diseases where there is an observed sex bias. Multiple IEI are due to loss-of-function variants on the X-chromosome, such as X-linked agammaglobulinemia and X-linked chronic granulomatous disease, and show much higher prevalence in males, as is expected for such X-linked recessive conditions given that males are hemizygous for the relevant gene¹²¹. In contrast, numerous common immune-mediated diseases, such as multiple sclerosis, systemic lupus erythematosus, and rheumatoid arthritis, exhibit strong sex bias, with higher prevalence in females compared to males^122–123. More generally, however, there is increased prevalence for IEI in females, particularly if X-linked recessive disorders are set aside¹²⁴. One could model genetic causes of disease and the presence of the X-chromosome as a genetic interaction. However, because some diseases do not manifest until later in life, an equally viable model would be the consideration of biological sex as an environmental factor. There is also an important stochastic component in X-linked disorders, namely X-inactivation. X inactivation is a process that occurs in females where one of the two X-chromosomes is randomly silenced early in embryonic development. The cell and all its descendants will continue to inactivate the same X chromosome. This process results in females being a mosaic of cells expressing either the maternal or paternal X chromosome. Skewing, in which the maternal X-chromosome is inactivated in significantly more or less than 50% of cells, is frequently seen in the general population¹²⁵. X-inactivation does not fully suppress gene expression of every gene on the X-chromosome (even setting aside the pseudoautosomal region that has a homologous region on the Y chromosome) and there is variability between individuals¹²⁶. In the context of common autoimmune diseases, a useful illustration of this gene dosage effect is the observation that male SLE patients are more likely to have two copies of the X-chromosome (Klinefelter’s syndrome)¹²⁷. In the context of IEI, the combination of variable X-inactivation and X-inactivation skewing can be an important contributor to variable expressivity and penetrance. A more comprehensive understanding of these interactions could yield valuable insights into disease mechanisms and contribute to the development of more effective diagnostic tools and therapeutic strategies for IEI.

The rise of single-cell sequencing approaches offers a novel opportunity to delve deeper into the connections between genotype and phenotype and the causes of phenotypic variation for a given genotype. For instance, by investigating the variations across single cells harboring different mutations, we can better evaluate on a molecular level the penetrance of these different mutations. If coupled with detailed understanding of the genetic or epigenetic factors also present, including for example a detailed examination of X-inactivation¹²⁸, it may be possible to better delineate the respective contributions of genetic, environmental, and stochastic factors in each case. This in turn could open up new avenues for disease prediction, prevention, and treatment.

Complex and somatic variants

Significant allelic and locus heterogeneity, including variability in the mode of inheritance, is seen in a number of IEI. Hyper IgE syndrome, a disorder characterized by frequent abscesses, severe eczema, frequent respiratory infections, and elevated IgE levels, is one such example. Hyper IgE was initially observed in kindreds with an autosomal pattern of inheritance in which the cause was ultimately found to be any of more than 100 dominant negative variants in STAT3^129–130. Hyper IgE syndrome can also be caused, however, in an autosomal recessive fashion due to homozygous variants in ZNF341^131–132, and there are both autosomal dominant and autosomal recessive forms driven by distinct mutations gp130 (encoded by IL6ST)^133–134. Autosomal recessive IEI can also be driven by compound heterozygosity in which two distinct deleterious alleles are inherited^135–137. Still more complicated would be the interaction of variants across a small number of genes. As the number of involved genes increases from one, comprehensive identification of the causal variants is increasingly difficult, though to develop computational tools to facilitate analysis of these so-called oligogenic disorders is underway¹³⁸. There is also a recent appreciation for the role of somatic variants in autoimmunity and IEI, with the most prominent example being VEXAS, recently described rare adult-onset syndrome that causes severe inflammatory and hematologic manifestations; In 2020, the NIH identified 25 men with somatic mutations affecting methionine-41 in UBA1, the major E1 enzyme that initiates ubiquitylation⁴². Of note, it is now clear that a subset of patients with the previously presumed polygenic autoimmune disease relapsing polychondritis in fact have VEXAS syndrome¹³⁹.

Since large language models intrinsically learn protein structure, they set themselves apart from other in silico methods in their applicability in predicting the function of variants. While other methods are typically trained on evolutionary constraints or allele frequencies, large language models are unsupervised and thus do not suffer from data circularity issues and can be broadly utilized to predict whether any mutation would be damaging to protein structure. These models also hold promise for examining complex variants, including compound heterozygotes, by factoring in multiple mutations simultaneously. Indeed, we and others have demonstrated that large language models can be effectively tailored to predict the effects of variants on different alternative spliced isoforms^67,68,70,140. Furthermore, these models can also be employed to functionalize short insertions and deletions (indels) with promising performance⁶³. These recent results underscore the immense potential of large language models in the field of genomics, ultimately driving our understanding of genetic diseases and accelerating the discovery of potential therapies.

Beyond genetics

Anti-cytokine autoantibodies have been identified in subsets of patients with SLE and other autoimmune diseases¹⁴¹. The discovery of autoantibodies against type-1 IFNs as a critical factor in determining the severity of COVID-19 further illustrate the potential for molecular phenotyping. This approach could be instrumental in identifying patient clusters with phenotypes that closely resemble an IEI. Indeed, the presence of autoantibodies against type-1 IFNs is not necessary but sufficient to cause critical COVID¹⁴². In a subsequent study following our initial exploration into autoantibodies COVID-19, we carried out a comprehensive profiling of COVID-19 patients at UCSF and healthy controls within San Francisco¹⁴³. Our findings echoed the initial discovery: approximately ~20% of patients with critical COVID-19 harbored autoAbs against type-1 IFNs, compared to a mere ~0.4% in the healthy control group. Further, our utilization of single-cell RNA-seq and surface protein profiling identified a deficient type-1 IFN response in the circulating immune cells of critically ill COVID-19 patients. In contrast, patients with mild to moderate symptoms demonstrated a significant upregulation of type-1 IFN-stimulated genes (ISGs), particularly in myeloid cells - a typical response to a viral infection. Interestingly, this response diminished as patients’ hospital stays extended. In critical cases, even at the earliest point of hospital interaction, typically during admission, a type-1 ISG was conspicuously absent. Moreover, we identified several surface proteins, with LAIR1 being notably prominent, displaying a reverse correlation with the type-1 IFN response. Indeed, the possibility exists that other autoantibodies, similar to type-1 IFN autoantibodies, may also play a role in mitigating antiviral responses¹⁴⁴ including the type1 ISG response¹⁴³.

Sequencing technologies

We envision that allelic series that are naturally or synthetically generated will be informative for understanding the molecular basis by which certain mutations give rise to IEI. Conventional next-generation sequencing is limited to reads of approximately 300 base pairs. In contrast, reads of tens to hundreds of kilo-base pairs can be achieved through the two primary long-read technologies, Pacific Biosciences’ SMRT (single molecule, real time) and Oxford Nanopore Technologies’ nanopore approach¹⁴⁵. Recent successes of these approaches include coverage of the heterochromatin regions of the human genome by the Telomere to Telomere consortium¹⁴⁶. To characterize haplotype-specific effects or to directly determine the sequence of a synthetic DNA molecule, long-read sequencing of single DNA molecules will be critically important. For natural variants, single molecule long-read sequencing allows us to identify all variants linked on a single haplotype, obviating the need for imputation from genotyping or sequencing data which may be challenging particularly for ultra rare variants. For synthetic variants, long-read sequencing has the potential to link specific-coding mutations with molecular phenotypes obviating the need for orthogonal barcoding which may be challenging due to recombination events that occur during viral production and packaging or challenges in synthesizing very long DNA molecules. Complementing the development of long-read sequencing are increases in throughput in single-cell genomics and more efficient molecular biology to capture DNA and mRNA information from cells^98,147. These approaches will be critically important for identifying rare somatic mutations and for studying the penetrance of disease causing mutations.

Disease-relevant cellular models

As discussed above, the effects of non-coding variants are frequently cell-type or other context specific, this is likely true for coding variants as well. Context specific effects extend beyond just cell type specificity but could be the interaction between environmental exposure and genetics or between developmental trajectory and genetics. Indeed, it will be important to develop cellular models that recapitulate the in vivo biology of immune cell development and differentiation or response to pathogens. There are exciting opportunities emerging to perturb circulating primary human immune cells ex vivo. However, to study the effects of variants on tissue resident cells or epithelial and mesenchymal cells, the differentiation of iPSC cells where candidate causal variants have been introduced will be needed. In these experiments, functional genomics can be utilized to separate the effects of reprogramming, differentiation, and the genetic perturbation. In neuroscience and other tissue biology, a number of efforts are already ongoing that leverages genetic multiplexing⁹⁹ that we developed to perform pooled differentiation experiments followed by single cell profiling called “cell villages”¹⁴⁸.

In vivo functionalization

The definitive proof for causality is usually conducted with in vivo experiments using mouse models. Some caution is necessary in translating the results of such studies to humans. Although the structure and function of the immune system is largely conserved between mice and humans, there are some areas of divergence^149,47. Moreover, for any specific coding variant it is important to consider not only whether that amino acid is conserved but also whether there is in fact a one-to-one homology between these genes between mouse and human. A recent study identifying an association between SLE and deleterious variants in P2RY8, an immunomodulatory GPCR that has been lost in mice, is a clear illustration of the need for such consideration¹⁵⁰. Even in this situation, of course, use of mouse-based experiments was nonetheless valuable, though it required the introduction of human P2RY8. Study of non-coding variants is likely to be still more difficult as it would require the insertion of large DNA regions into the germline, though approaches relevant to such investigation have certainly been developed¹⁵¹. For example, non-coding variants of ERAP2 are associated with Crohn’s disease and are known to modify the expression of different isoforms in vitro but the gene is absent in mouse¹⁵². While we have studied the effects of different ERAP2 isoforms by inserting them into the mouse genome, the exact context specific effects of the non-coding sequences remain difficult to study in vivo as it would require the insertion of large DNA sequences through bacterial artificial chromosomes (BACs). Non-human primates may indeed be intriguing model organisms for functionalizing causal human variants for inborn errors of immunity¹⁵³.

Future opportunities for clinical applications

Diagnoses

Clinical diagnoses start with phenotyping of patients. When patients are suspected of having a Mendelian disorder, exome and genome sequencing can be performed to identify potential genetic causes. We believe functional genomics has an important role to play in both powering phenotyping of patients and annotating genetic variants which together can improve the diagnoses of IEI or those disorders that phenocopy an IEI. For genetic variants, the vast majority of variants in an individual will still be annotated as variants of unknown significance. Our work shows that it is now feasible to correctly identify a disease causing coding variant utilizing large-language models. We have also shown that single-cell sequencing may be helpful in phenotyping patients, some of whom may not have a germline mutation but may have somatic mutations or other large effect immunological lesions that result in phenotypes that resemble those with an IEI. As an example, we have utilized single-cell RNA-sequencing as an approach to phenotype patients coming into the UCSF clinic presenting atypical immunological symptoms. We collected PBMCs from 15 samples and performed multiplexed single-cell RNA-sequencing while the patients also received a standard clinical workup. We then organized a Hackathon, pairing computational biologists, basic immunologists, and clinicians to assign the single-cell transcriptomic profile to each clinical profile. The six groups that participated in this exercise correctly categorized ~70% of the cases.

Targeted biologic treatments

The development of advanced characterization techniques for diseases at the genetic and molecular levels has ushered in a new era of personalized medicine, significantly expanding the therapeutic landscape. This is particularly relevant for rare genetic diseases such as IEI, which have often been neglected in drug development due to their low prevalence. Conversely, knowledge of manifestations of IEI can shed light on expected side effects from targeted manifestations. For example, the cytotoxic T-lymphocyte antigen-4 (CTLA-4) protein is an important inhibitory regulator of immune responses. By competing for binding to CD80 and CD86, it is functionally an antagonist of the co-stimulatory molecule CD28. Whereas homozygous ablation of CTLA-4 was long known to result in fatal autoimmunity in mice^154–155, individuals with immunodysregulation with heterozygous CTLA-4 variants were more recently identified^156–157, a syndrome termed CTLA-4 haploinsufficiency with autoimmune infiltration (CHAI). Clinical presentation can include both immunodeficiency (hypogammaglobulinemia in the majority of patients) and autoimmunity such as cytopenias, thyroiditis, arthritis, colitis, type 1 diabetes, or pneumonitis^158–159. This is very similar to the side effects that have been observed following infusion of CTLA-4 blocking antibody ipilimumab¹⁶⁰. Conversely, Abatacept, a CTLA-4 Ig that is approved in the treatment of rheumatoid arthritis and juvenile idiopathic arthritis, has been repurposed to treat patients with CHAI with some success¹⁶¹. Recently, a gain-of-function mutation in the STAT4 gene was identified as the cause behind debilitating pansclerotic morphea. This groundbreaking finding prompted a trial of treatment with the JAK inhibitor Ruxolitinib in the affected patients, which led to the resolution of the clinical symptoms in treated patients¹⁶².

Therefore, the comprehensive characterization of IEI not only opens avenues for repurposing existing therapies but also helps understand and anticipate potential adverse effects for some biologic and targeted treatments.

Retroviral gene replacement therapy

While the aforementioned process has been a success for some IEI, treatment continues to rely on aggressive and high risk procedures for many other patients, which has created the need for alternative therapies. Severe combined immunodeficiency (SCID), in which there are defects in both cellular and humoral adaptive immunity, has historically required treatment with allogeneic hematopoietic stem cell transplantation (HSCT). While allogeneic HSCT can be curative, it is associated with difficulties identifying HLA-matched donors, and significant morbidity and mortality associated with conditioning regimen¹⁶³, risk of graft failure, and graft versus host disease¹⁶⁴. This has led to the development of gene editing technologies, which efficiently enables the correction of the genetic defect while preserving other cell functions. In recent years, however, several groups have used lentiviral delivery platforms to introduce a wild-type gene into the patient’s own HSCs which are then used for autologous HSCT. This approach has been successfully applied to multiple genetic etiologies of SCID, including SCID-X1 (hemizygous mutation in the common γ chain¹⁶⁵), ADA-SCID (homozygous mutation in adenosine deaminase¹⁶⁶), and Artemis-SCID (homozygous mutation in DCLRE1C-encoded DNA repair protein Artemis¹⁶⁷). Early attempts using γ-retrovirus as the delivery platform were successful but halted after high rates of insertional oncogenesis^168–169. Subsequent modifications have demonstrated improved safety and have been approved for treatment of ADA-SCID¹⁷⁰. Beyond SCID, other PIDD in which this approach is being applied include Wiskott-Aldrich syndrome¹⁷¹, X-linked chronic granulomatous disease¹⁷², and leukocyte adhesion deficiency¹⁷³. It is still early days for this therapy and continued study is needed to better understand the benefits of different conditioning regimens as well as long-term safety and efficacy.

Antisense oligos for haploinsufficient disease

Antisense oligonucleotides (ASOs) represent a promising therapeutic approach to treat genetic disorders characterized by haploinsufficiency, a condition where a person has only one functional copy of a gene, and that one copy does not produce enough gene product (typically a protein) to exhibit normal function. This can lead to various diseases, depending on the specific gene involved. ASOs are short, synthetic pieces of DNA or RNA that can bind to the messenger RNA (mRNA) produced by a specific gene. Once an ASO is bound to an mRNA molecule, it can influence gene expression in several ways.

For haploinsufficiency, ASOs can be designed to increase the expression of a gene. They work by binding to regulatory regions on the mRNA or pre-mRNA, blocking the binding of repressors or facilitating the binding of activators, thereby increasing the amount of protein produced. For genes affected by haploinsufficiency, this approach could effectively increase the gene product to a normal level. Another promising ASO approach for some haploinsufficient conditions involves ‘exon skipping’. This strategy can be especially useful for diseases like Duchenne Muscular Dystrophy, where specific gene mutations result in a dysfunctional protein. In this context, the ASO is designed to bind to the mRNA at the location of the mutation, causing the cellular machinery to ‘skip’ over the mutation during protein synthesis. The resulting protein, while not identical to the fully functional protein, is often still able to perform its main functions, leading to improved patient outcomes.

However, while ASOs hold great promise, it is important to note that this field is still in its relative infancy. Many challenges exist in the design, delivery, and specificity of these molecules, and further research is needed to ensure their safe and effective use in patients. The use of ASOs for the treatment of haploinsufficiency is an area of active research, and advancements in this field could open up new possibilities for treating a range of genetic disorders.

Cell based therapies

Finally, cell based therapies are emerging as a promising pillar of modern medicine. Specifically, Chimeric Antigen Receptor T-cell (CAR-T) therapy, has gained significant traction since the early trials in B-cell acute lymphoblastic leukemia, where it led to remissions in patients who had been failing multiple lines of prior treatments^174.175. CAR-Ts can be generated by transducing T cells with retroviral or lentiviral vectors expression the CAR construct, or more recently, using CRISPR/Cas9 to the TRAC locus¹⁷⁶. These therapies involve using living cells to treat or potentially cure diseases, often by bolstering or replacing parts of the immune system. Although initially developed to treat liquid and now solid tumors, in vivo cell based therapies with high degree of tropism might be particularly exciting for treating IEI. For example, In CAR-T therapy, T cells are taken from a patient’s blood. In a lab, a gene is added to the T cells that allows them to produce receptors (CARs) on their surface. These receptors enable the T cells to recognize and attach to specific proteins on tumor cells. The modified cells are then multiplied in the lab and injected back into the patient, where they can seek out and destroy cancer cells.

Applying similar principles, CAR-T therapy could potentially be adapted for treating IEI. For example, recent work has shown that engineer T cells with CARs can recognize and attach to B cells producing autoantibodies, which are a major driver of some autoimmune conditions including SLE¹⁷⁷. Upon infusion back into the patient, these CAR-T cells could selectively target and eliminate the autoreactive B cells, thereby reducing the autoimmune response. It is important to note that this approach would need to be carefully controlled to avoid depleting B cells entirely, as they also play important roles in fighting infections. Such a targeted approach could offer significant benefits over current treatments, which often involve general immunosuppression that can leave patients susceptible to infections. Caution also needs to be exercised to monitor or mitigate chromosomal abnormalities that may be introduced by the gene editing event. We have recently described the presence of aneuploidy (3–20%) in edited T cells and a strategy for mitigating these effects¹⁷⁸. However, this field is still emerging, and further research is required to ensure the safety and efficacy of such treatments, particularly in managing the potential off-target effects and other complications. Nonetheless, the potential of CAR-T and other cell-based therapies to treat IEI and related autoimmune diseases offers an exciting direction for future research and clinical practice.

Conclusions

A comprehensive understanding of gene functions and their impact on immune response is crucial for the development of targeted therapies that can possibly mitigate genetic anomalies associated with these disorders. Thus, functional genomics may play a transformative role in shaping our approach to IEI, from diagnosis and treatment to advancing our overall management of these conditions.